Pith. sign in

REVIEW 5 major objections 5 minor 1 cited by

VMoGE—a variational mixture of graph neural experts—claims that separating EEG into four frequency bands and giving each band its own graph-structured expert, with a variational gate that learns per-patient weights, yields better dementia c

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

VMoGE, a variational mixture of per-frequency-band graph experts, reports AUC up to 0.89 for Alzheimer's vs. healthy EEG and links learned band weights to known dementia markers.

T0 review reviewed 2026-08-04 challenge →

load-bearing objection New combination of per-band GMRF-prior experts and variational MoE for EEG dementia classification, but the unstated subject-level cross-validation is the load-bearing weakness; deserves revision, not rejection. the 5 major comments →

arxiv 2510.11917 v3 pith:4BYSMZGC submitted 2025-10-13 cs.LG

Variational Mixture of Graph Neural Experts for Alzheimer's Disease Recognition across Frequency Bands in EEG Brain Networks

Jun-En Ding , Anna Zilverstand , Shihao Yang , Albert Chih-Chieh Yang , Feng Liu This is my paper
classification cs.LG
keywords Alzheimer's diseaseEEGmixture of expertsgraph neural networksvariational inferenceGaussian Markov random fieldfrontotemporal dementiafrequency bands
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper is trying to establish that a model which treats each EEG frequency band (delta, theta, alpha, beta) as a separate expert in a variational mixture-of-experts graph network can diagnose and stage dementia more accurately than full-band EEG classifiers while remaining interpretable. On two EEG datasets, it reports AUC improvements of between 4 and 10 percentage points over comparison models, with an AUC of 0.89 for healthy controls versus Alzheimer's disease in the main comparison. It further claims that the learned gating weights shift toward slow-wave bands as cognition declines, correlate with MMSE scores and CDR severity, and produce spatial maps consistent with known AD pathology. If true, this would make EEG a more practical and clinically legible tool for distinguishing Alzheimer's disease from frontotemporal dementia and for tracking disease progression.

Core claim

VMoGE's central claim is that frequency-specific graph structure, not just frequency-specific features, is what lets EEG separate dementia subtypes and stages. The model builds four separate graphs—one per band—with 19 EEG channels as nodes, models each band's latent representation under a Gaussian Markov random field prior whose precision is derived from the band graph, and routes inputs through a variational gating network that learns per-sample weights for combining the four experts' predictions. The paper reports that this design reaches AUC 0.89 for HC vs AD and 0.78 for HC vs FTD on the Open AD dataset, and 0.65-0.73 across CDR staging tasks on a session-based dataset, consistently abo

What carries the argument

The load-bearing mechanism is the closed-form KL divergence between a variational posterior and a Gaussian Markov random field prior: D_KL = 1/2 [tr(QΣ) + μ^T Q μ − C − log|Σ| + log|Q|]. This term injects band-specific graph topology into the training objective; the precision matrix Q(k) = I − D^{−1/2} A(k) D^{−1/2} encodes which channels should have similar latent representations in that band. Around this sit four expert variational graph encoders (one per band) and a gating network π(k)(H') = softmax_k(w_k^T φ(H')) that turns the concatenated band features into a per-sample weighted average of expert logits. The multi-granularity transformer (MGT-NFE) supplies node features, using 1-D conv

Load-bearing premise

The load-bearing premise is that the GMRF prior's precision matrix is invertible so the KL term (and log|Q|, Q^{-1}) is well-defined; as written Q is the normalized graph Laplacian, which is singular for every connected graph, so the stated math relies on an unstated shift or pseudo-inverse (the paper does test shifted variants in Table III, but not in the main equations).

What would settle it

Run the reported training procedure with Q exactly equal to I − D^{−1/2} A D^{−1/2} on the 19-channel 10-20 graph used in both datasets. Since that matrix has a zero eigenvalue, log|Q| and Q^{-1} cannot be computed, so the run either fails or forces an implicit regularization; whichever shifted Q is used, the AUC numbers should be re-reported. A reproduction that changes ε in Q + εI and sees the HC-vs-AD AUC move outside the reported ±0.07 would show the result is tied to an unstated detail rather than to the GMRF structure itself.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • If the reported AUCs hold, a 19-channel resting EEG can separate AD, FTD, and healthy aging at a level (AUC 0.78-0.89) that plausibly supports screening and differential diagnosis.
  • The gating weights give a per-patient, per-band score that tracks MMSE and CDR, so the same model output could be used as a continuous severity index rather than only a binary label.
  • The spatial maps imply that the signal for AD is posterior alpha/theta slowing while FTD vs AD turns on central/temporal beta changes; any full-band model would blur those distinct signatures.
  • The GMRF prior's largest benefit appears in the small-sample staging tasks (CDR=0 vs CDR=2 AUC 0.77 to 0.81), indicating graph-structure regularization is most valuable when training data are scarce.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Because the gating weights already correlate with MMSE, a natural extension is to train VMoGE on longitudinal EEG from the same patients and ask whether shifts in delta/theta weight precede the MMSE decline; the paper does not report such a test.
  • The paper's own frequency-band ablations suggest a cheap test: an AD-vs-HC classifier built only from the delta and theta experts should nearly match the full model, since those bands carry most of the slow-wave signal; the paper does not isolate this two-band model.
  • The precision matrix as written is singular for any connected graph, so the experiments must in practice use one of the shifted variants (L+λI or Lnorm+λI) from Table III; making that choice explicit, and testing how much the final AUC depends on ε, would settle whether the reported gains are tied to an unstated regularization detail.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. The paper proposes VMoGE, a variational mixture-of-experts framework for EEG-based dementia classification. Each frequency band (δ, θ, α, β) is processed by a multi-granularity transformer to obtain node features, then encoded by a variational GCN whose latent variables are regularized toward a Gaussian Markov random field prior built from an adjacency matrix. A learned gating network weights the four band experts, and the final prediction is a weighted mixture of expert logits. The authors report five-fold cross-validated AUC/ACC on two datasets (Open AD, Session-based AD) for three pairwise tasks each, claiming +4% to +10% AUC over state-of-the-art methods, with the headline result AUC 0.89 for HC vs AD on the Open AD dataset. They also present ablation studies for temporal granularities, λ regularization strength, and single-expert vs full-expert variants, and interpret learned gating weights as biomarkers correlated with MMSE, age, and CDR severity.

Significance. If the reported empirical results are trustworthy, VMoGE would be a meaningful contribution to EEG-based dementia diagnosis and staging, with a plausible design that combines frequency-specific graph priors, variational inference, and mixture-of-experts gating. The paper is also unusually thorough in its ablations: temporal granularity configurations, λ sensitivity, and single-expert comparisons are all examined, and the interpretability analysis targets clinically relevant variables. However, the central empirical claim rests on several load-bearing assumptions that are not verified as written: the cross-validation split is not stated to be subject-level, the GMRF prior in the main derivation is singular for the normalized Laplacian and the adjacency matrix A(k) is never defined, and the main results table disagrees with the ablation table for the same configuration. Because these issues directly affect the validity and reproducibility of the claimed AUC gains, the paper cannot be accepted in its present form.

major comments (5)
  1. [§V.A–V.B, Table I] The evaluation uses five-fold cross-validation, but the paper never states that folds are split by subject. Both datasets contain multiple epochs per subject: Open AD has 12–14 min recordings (88 subjects) and Session-based AD has three 10-s segments per subject (123 subjects). If epochs from the same subject appear in both training and test folds, the model can memorize subject-specific signatures, inflating AUC. This is the most serious issue: it directly undermines the claimed +4% to +10% improvements. Please specify the exact split strategy (e.g., GroupKFold at subject level) and, if epoch-level splits were used, re-run all experiments with subject-level grouping.
  2. [§III.C, Eq. (9)–(10), Eq. (26)] The precision matrix Q(k) is defined as the symmetric normalized Laplacian I − D^{-1/2} A D^{-1/2}, which is singular for any connected graph. Equation (10) requires Q^{-1} and Eq. (26) requires log|Q|, both of which are undefined for a singular precision matrix. Table III later considers GMRF (L+λI) and GMRF (Lnorm+λI), showing the authors are aware of regularized variants, but the main derivation does not state which one is used. The paper must specify the actual regularized precision matrix (e.g., Q + εI), a pseudo-inverse, or a constrained GMRF formulation.
  3. [§III.C, Eq. (9), §III.D, Eq. (11)] The adjacency matrix A(k) is never defined. The method depends on A(k) for the GMRF prior, for the GCN encoder via Â(k), and for the KL divergence. No formula, correlation measure, thresholding rule, or data-driven construction is given. Without this, the graph structure is underspecified and the experiments are not reproducible. Please provide the exact construction of A(k) for each frequency band.
  4. [Table I vs Table IV] The reported VMoGE results are internally inconsistent. In Table I, VMoGE achieves AUC 0.89 ± 0.07 for HC vs AD (Open AD) and 0.79 ± 0.07 for FTD vs AD; in Table IV, the same configuration is reported as 0.92 ± 0.10 and 0.87 ± 0.12. The text cites the Table I value (0.89) in Section V.C but Section V.D's ablation cites 0.92. The authors must clarify which table corresponds to the final model, how random seeds and hyperparameter selection were handled, and why the values differ.
  5. [§VI.B, Fig. 5] The biomarker correlations are computed between learned gating weights and MMSE/CDR on the same datasets and labels used for training. These are post-hoc correlations, not out-of-sample evidence, and no multiple-comparison correction is reported across the many band×task×clinical-variable tests. Some of the reported p-values (e.g., r = −0.336, p = 0.0226) may not survive correction. Please clarify whether these correlations are computed on held-out data and whether any correction for multiple testing was applied.
minor comments (5)
  1. [Title/Abstract] The abstract title uses “Recognition across Frequency Bands in EEG Brain Networks,” while the full-text title reads “Biomarker Recognition in EEG Brain Networks.” This inconsistency should be resolved.
  2. [Eq. (20)] The ELBO writes a single KL term but the model has K experts; the objective should make the summation over k explicit, otherwise the role of λ and the per-band priors is ambiguous.
  3. [§V.D, Table III] The text says “GMRF methods . . . pure GMRF excels at both extremes, reaching the highest score with 0.84 at λ=0.8 for FTP vs AD.” “FTP” is a typo for “FTD,” and the table should be checked for similar notation errors.
  4. [References] References [28] and [60] are the same arXiv paper (GraphDIVE / “Graph classification by mixture of diverse experts”); the duplicate should be removed or one citation should be used throughout.
  5. [General presentation] The paper does not include a limitations subsection. Given the small sample sizes and the exploratory nature of the biomarker analysis, a brief discussion of limitations (e.g., no statistical significance testing between model AUCs, no confidence intervals for AUC differences) would strengthen the manuscript.

Circularity Check

3 steps flagged

Reported AUC gains are partly selected from test-task hyperparameter grids and the evaluation pipeline does not enforce subject-level splits; biomarker correlations are post-hoc on training data.

specific steps
  1. fitted input called prediction [Section V.C–V.D, Table III vs Table I]
    "Table III presents the AUC scores for λ values ranging from 0.1 to 1.0 across three types of binary classification tasks on both Open and Session-based AD datasets, with red highlighting indicating the optimal λ value for each prior type within each task. ... In overall comparison, VMoGE achieved AUC = 0.78, ACC = 0.74 in HC vs FTD (Open dataset) ... In HC vs AD (Session-based AD), our method achieved AUC = 0.89, ACC = 0.83, outperforming EEGNet (AUC = 0.81, ACC = 0.76) by +9.9% and +9.2%, respectively."

    The final VMoGE results are reported after selecting λ per task from the test-task AUC grid in Table III (and the granularity configuration from Fig. 3), with no separate validation set described. The reported AUC is therefore the best of the tuned configurations evaluated on the same task labels, so the +4% to +10% AUC improvement is a selection statistic rather than an out-of-sample prediction.

  2. other [Section III.A and Section V.A–V.B (epoch sampling and five-fold CV)]
    "The EEG data were first segmented into non-overlapping epochs ... we employ two diverse datasets using five-fold cross-validation ... recording durations averaging 12-14 minutes per subject ... three 10-second segments free of significant noise were extracted from the closed-eye phase of the raw EEG data."

    Instances are epochs, not subjects; with 12–14 min recordings per subject (Open AD) and three segments per subject (Session-based), and no statement that folds are grouped by subject, the same subject's epochs can appear in both training and test folds. The classifier can then achieve high AUC by recognizing subject-specific EEG signatures rather than disease, so the reported test performance is not independent of the training input for those subjects.

  3. fitted input called prediction [Abstract and Section VI.B]
    "the expert gating weights correlate with MMSE scores and CDR severity ... In HC vs AD classification, the δ-band exhibited a significant negative correlation (r=−0.336,p= 0.0226)."

    The gating weights are learned on these same subjects' diagnostic labels, and MMSE/CDR are strongly correlated with those labels. Reporting correlations of trained weights with MMSE/CDR on the training subjects is a post-hoc fit of the model's own allocation to clinical variables correlated with its supervision, not an independent validation of the biomarker.

full rationale

VMoGE's core model is not equation-level circular: the classification objective (Eq. 27) is trained against external labels and the reported AUC is a measured quantity, so the network is not defined in terms of its output. The circularity is in the evaluation/interpretation chain. First, the paper tunes λ (Table III) and granularity (Fig. 3) per task on the same test-task tables and then reports the resulting AUC as VMoGE performance; this makes the headline +4–10% improvement a selected maximum, i.e., a fitted value presented as a prediction. Second, the described pipeline uses epochs as instances from long multi-epoch recordings while specifying only 'five-fold cross-validation' without subject-level grouping; under that description, the same subject's epochs can be in both training and test folds, so the reported generalization can reduce to subject-identity memorization. Third, the interpretability claims in Sec. VI are correlations of trained gating weights with MMSE/CDR on the same subjects used for training, so they are partly restatements of the training labels' correlates. The singular GMRF precision matrix (Eqs. 9–10, 26) is a genuine mathematical flaw but is an internal inconsistency, not circularity, and it does not by itself affect the circularity score. Self-citations (e.g., dataset [39]) are not load-bearing. Overall, the central performance claim is compromised by test-set selection and potential subject leakage, so a score of 6 is warranted.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 0 invented entities

The central empirical result rests on under-specified graph constructs: the adjacency A(k) is never defined, and the GMRF prior built from it is mathematically ill-defined as written. The per-task selection of λ and granularity adds fitting degrees of freedom, and the biomarker interpretations are post-hoc rather than pre-registered.

free parameters (4)
  • KL weight λ = varies by task/dataset; e.g., 0.6-0.8 for Session CDR=0 vs CDR=2, 0.1 for some Open AD tasks
    Table III scans λ ∈ {0.1, 0.2, 0.6, 0.8, 1.0} and reports the best AUC per task; no held-out validation for this selection.
  • Temporal granularity configuration = e.g., Medium for Open AD HC vs AD; Mixed-2 and Single-Fine for other Open tasks; Coarse for Session CDR=0 vs CDR=2
    Fig. 3 selects among Fine/Medium/Coarse/Mixed-1/Mixed-2 based on F1; this discrete hyperparameter is tuned per task.
  • Graph adjacency A(k) = not reported
    The edges of each band-specific graph are never defined by an equation, fitted value, or reference; the entire GCN and GMRF prior depend on this unreported modeling choice.
  • Network hyperparameters (transformer layers, latent dim, GCN/MLP widths) = not reported
    No values or search ranges are given for Ltr, dz, GCN layers, or MLP sizes, so the reported results cannot be audited or reproduced from the text.
axioms (4)
  • ad hoc to paper Q(k) = I - D^{-1/2} A(k) D^{-1/2} is a valid precision matrix for a proper Gaussian prior N(0, Q^{-1})
    The normalized Laplacian is singular for connected graphs, so Q^{-1} and log|Q| in Eqs. 10 and 26 are undefined. The paper never states a regularization or pseudo-inverse.
  • domain assumption A(k) is a known, meaningful connectivity matrix for each frequency band, with edges (i,j) iff Q(k)_ij != 0
    No equation or reference defines how A(k) is computed from the RBP features; both the GCN encoder and the GMRF prior rely entirely on this matrix.
  • standard math Mean-field factorized Gaussian posterior sufficiently approximates the latent band graphs
    Eq. 12 assumes a factorized Gaussian posterior over nodes, standard in variational graph autoencoders, but no analysis is given for EEG graph structure.
  • domain assumption Relative band power features from Welch's method contain the discriminative information for AD/FTD classification
    All input features are RBP in four bands; any limitation of this spectral representation propagates into the central classification claims.

reviewed 2026-08-04 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Variational Mixture of Graph Neural Experts for Alzheimer's Disease Recognition across Frequency Bands in EEG Brain Networks." pith.science (2026). https://pith.science/paper/4BYSMZGC

@misc{pith2026251011917,
  author       = {Pith},
  title        = {Pith review of: Variational Mixture of Graph Neural Experts for Alzheimer's Disease Recognition across Frequency Bands in EEG Brain Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4BYSMZGC}},
  note         = {Machine review of arXiv:2510.11917}
}
Share X Bluesky LinkedIn Reddit HN
abstract

Dementia disorders such as Alzheimer's disease (AD) and frontotemporal dementia (FTD) exhibit overlapping electrophysiological signatures in EEG that challenge accurate diagnosis. Existing EEG-based methods are limited by full-band frequency analysis, which hinders the precise differentiation of dementia subtypes and severity stages. To address this limitation, we propose a Variational Mixture of Graph Neural Experts (VMoGE) framework that integrates multi-band EEG analysis with variational graph neural networks and a mixture-of-experts architecture. Each expert specializes in a specific EEG frequency band and models brain connectivity using a Gaussian Markov random field prior, while a variational gating mechanism adaptively integrates the expert outputs. This design enables the model to learn frequency-specific brain network representations while modeling latent uncertainty through variational inference. Experimental results on two EEG dementia datasets show that VMoGE achieves strong performance, with an AUC of 0.89 for HC vs. AD classification in the main comparison and competitive results across dementia subtyping and CDR staging tasks. Clinically, VMoGE offers three key translational values: the expert gating weights correlate with MMSE scores and CDR severity; slow-wave $\delta$- and $\theta$-band contributions are associated with AD-related EEG slowing and disease progression; and spatially localized activation maps reveal posterior $\theta$- and $\alpha$-band alterations and region-specific $\beta$-band changes, providing neurophysiologically interpretable markers aligned with known AD neuropathology.

Figures

Figures reproduced from arXiv: 2510.11917 by Albert Chih-Chieh Yang, Anna Zilverstand, Feng Liu, Jun-En Ding, Shihao Yang.

Figure 1
Figure 1. Figure 1: Diagram of MGT-NFE for node feature extraction, [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the VMoGE framework for AD biomarker identification and prediction. VMoGE framework first extracts node features from multi-channel EEG signals (channels C1–C19) by integrating spatial and frequency band features through a 1D-CNN and FFT-based MGT-NFE module. The prior graph structure is constructed using a GMRF, and a variational router models the latent distribution to capture structural corr… view at source ↗
Figure 3
Figure 3. Figure 3: Ablation study of MGT-NFE module using different granularity components for EEG feature extraction. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Box plots showing the distribution of mixture weights [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The expert gating weight analysis for different dementia indicator analysis. The top row shows age-based comparisons, [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Diagram of electrode spatial positions and correspond [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Combined visualization of topographical maps (upper) and heatmaps (lower), showing spatial distributions and activation [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Technological Advances in Detecting and Managing Cognitive Impairment in Older Adults: Trends, Challenges, and Future Directions

    cs.LG 2026-07 conditional novelty 3.0

    Narrative review of cognitive-impairment detection technologies concludes that reported accuracies are often inflated by weak validation and that progress depends on multimodal, longitudinally validated, externally te...

Reference graph

Works this paper leans on

80 extracted references · 8 linked inside Pith · cited by 1 Pith paper

  1. [1]

    2025 alzheimer’s disease facts and figures,

    “2025 alzheimer’s disease facts and figures,”Alzheimer’s & Dementia, vol. 21, p. e70235, Apr. 2025

    2025

  2. [2]

    Mohw announces latest epidemiological survey results on dementia in taiwan communities

    Ministry of Health and Welfare, Taiwan, “Mohw announces latest epidemiological survey results on dementia in taiwan communities.” https://www.mohw.gov.tw/cp-16-78102-1.html, 2024. Accessed: 2025- 08-30

    2024

  3. [3]

    Frontotemporal dementias: a review,

    N. D. Weder, R. Aziz, K. Wilkins, and R. R. Tampi, “Frontotemporal dementias: a review,”Annals of general psychiatry, vol. 6, no. 1, p. 15, 2007

    2007

  4. [4]

    Behaviour in frontotemporal dementia, alzheimer’s disease and vascular dementia,

    D. Bathgate, J. Snowden, A. Varma, A. Blackshaw, and D. Neary, “Behaviour in frontotemporal dementia, alzheimer’s disease and vascular dementia,”Acta neurológica scandinavica, vol. 103, no. 6, pp. 367–378, 2001

    2001

  5. [5]

    Differences in multimodal electroencephalogram and clinical correlations between early-onset alzheimer’s disease and frontotemporal dementia,

    N. Lin, J. Gao, C. Mao, H. Sun, Q. Lu, and L. Cui, “Differences in multimodal electroencephalogram and clinical correlations between early-onset alzheimer’s disease and frontotemporal dementia,”Frontiers in Neuroscience, vol. 15, p. 687053, 2021. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 13

    2021

  6. [6]

    Neural biomarker diagnosis and prediction to mild cognitive impairment and alzheimer’s disease using eeg technology,

    B. Jiao, R. Li, H. Zhou, K. Qing, H. Liu, H. Pan, Y . Lei, W. Fu, X. Wang, X. Xiao,et al., “Neural biomarker diagnosis and prediction to mild cognitive impairment and alzheimer’s disease using eeg technology,” Alzheimer’s research & therapy, vol. 15, no. 1, p. 32, 2023

    2023

  7. [7]

    Eeg biomarkers in alzheimer’s and prodromal alzheimer’s: A comprehensive analysis of spectral and connectivity features,

    C. A. Chetty, H. Bhardwaj, G. P. Kumar, T. Devanand, C. A. Sekhar, T. Aktürk, I. Kiyi, G. Yener, B. Güntekin, J. Joseph,et al., “Eeg biomarkers in alzheimer’s and prodromal alzheimer’s: A comprehensive analysis of spectral and connectivity features,”Alzheimer’s Research & Therapy, vol. 16, no. 1, p. 236, 2024

    2024

  8. [8]

    An explainable and efficient deep learning framework for eeg-based diagnosis of alzheimer’s and frontotemporal dementia,

    W. Khan, M. S. Khan, S. N. Qasem, W. Ghaban, F. Saeed, M. Hanif, and J. Ahmad, “An explainable and efficient deep learning framework for eeg-based diagnosis of alzheimer’s and frontotemporal dementia,” Frontiers in Medicine, vol. 12, p. 1590201, 2025

    2025

  9. [9]

    Diagnose alzheimer’s disease and mild cogni- tive impairment using deep cascadenet and handcrafted features from eeg signals,

    K. Rezaee and M. Zhu, “Diagnose alzheimer’s disease and mild cogni- tive impairment using deep cascadenet and handcrafted features from eeg signals,”Biomedical Signal Processing and Control, vol. 99, p. 106895, 2025

    2025

  10. [10]

    Olfactory eeg based alzheimer disease classification through transformer based feature fusion with tunable q-factor wavelet coefficient mapping,

    B. Cansiz, H. ILHAN, N. Aydin, and G. Serbes, “Olfactory eeg based alzheimer disease classification through transformer based feature fusion with tunable q-factor wavelet coefficient mapping,”Frontiers in Neuroscience, vol. 19, p. 1638922, 2025

    2025

  11. [11]

    Deep ensemble learning with transformer models for enhanced alzheimer’s disease detection,

    S. Latif, N. U. Islam, Z. Uddin, K. M. Cheema, S. S. Ahmed, and M. F. Khan, “Deep ensemble learning with transformer models for enhanced alzheimer’s disease detection,”Scientific Reports, vol. 15, no. 1, p. 24720, 2025

    2025

  12. [12]

    Investigating convolutional and transformer-based models for classifying mild cognitive impairment using 2d spectral images of resting-state eeg,

    M. ¸ Seker and M. S. Özerdem, “Investigating convolutional and transformer-based models for classifying mild cognitive impairment using 2d spectral images of resting-state eeg,”Biomedical Signal Pro- cessing and Control, vol. 105, p. 107667, 2025

    2025

  13. [13]

    Dice-net: a novel convolution-transformer architecture for alzheimer detection in eeg signals,

    A. Miltiadous, E. Gionanidis, K. D. Tzimourta, N. Giannakeas, and A. T. Tzallas, “Dice-net: a novel convolution-transformer architecture for alzheimer detection in eeg signals,”IEEe Access, vol. 11, pp. 71840– 71858, 2023

    2023

  14. [14]

    Lead: Large foundation model for eeg-based alzheimer’s disease detection,

    Y . Wang, N. Huang, N. Mammone, M. Cecchi, and X. Zhang, “Lead: Large foundation model for eeg-based alzheimer’s disease detection,” arXiv preprint arXiv:2502.01678, 2025

    arXiv 2025

  15. [15]

    A dual path graph neural network framework for dementia diagnosis,

    D. Zhang and C. Zhu, “A dual path graph neural network framework for dementia diagnosis,”Scientific Reports, vol. 15, no. 1, p. 23319, 2025

    2025

  16. [16]

    A multi-graph convolutional network method for alzheimer’s disease diagnosis based on multi- frequency eeg data with dual-mode connectivity,

    Q. Xu, L. An, H. Yang, and K.-S. Hong, “A multi-graph convolutional network method for alzheimer’s disease diagnosis based on multi- frequency eeg data with dual-mode connectivity,”Frontiers in Neuro- science, vol. 19, p. 1555657, 2025

    2025

  17. [17]

    A novel graph neural network method for alzheimer’s disease classifica- tion,

    Z. Zhou, Q. Wang, X. An, S. Chen, Y . Sun, G. Wang, and G. Yan, “A novel graph neural network method for alzheimer’s disease classifica- tion,”Computers in Biology and Medicine, vol. 180, p. 108869, 2024

    2024

  18. [18]

    Eeg-based brain functional network analysis for differential identification of dementia- related disorders and their onset,

    A. T. Adebisi, H.-W. Lee, and K. C. Veluvolu, “Eeg-based brain functional network analysis for differential identification of dementia- related disorders and their onset,”IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 32, pp. 1198–1209, 2024

    2024

  19. [19]

    Adap- tive gated graph convolutional network for explainable diagnosis of alzheimer’s disease using eeg data,

    D. Klepl, F. He, M. Wu, D. J. Blackburn, and P. Sarrigiannis, “Adap- tive gated graph convolutional network for explainable diagnosis of alzheimer’s disease using eeg data,”IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 31, pp. 3978–3987, 2023

    2023

  20. [20]

    Variational graph normalized autoencoders,

    S. J. Ahn and M. Kim, “Variational graph normalized autoencoders,” in Proceedings of the 30th ACM international conference on information & knowledge management, pp. 2827–2831, 2021

    2021

  21. [21]

    Dynamic causal explanation based diffusion-variational graph neural network for spatiotemporal forecasting,

    G. Liang, P. Tiwari, S. Nowaczyk, S. Byttner, and F. Alonso-Fernandez, “Dynamic causal explanation based diffusion-variational graph neural network for spatiotemporal forecasting,”IEEE Transactions on Neural Networks and Learning Systems, 2024

    2024

  22. [22]

    A survey on mixture of experts in large language models,

    W. Cai, J. Jiang, F. Wang, J. Tang, S. Kim, and J. Huang, “A survey on mixture of experts in large language models,”IEEE Transactions on Knowledge and Data Engineering, 2025

    2025

  23. [23]

    Dynamic modeling of patients, modalities and tasks via multi-modal multi-task mixture of experts,

    C. Wu, Z. Shuai, Z. Tang, L. Wang, and L. Shen, “Dynamic modeling of patients, modalities and tasks via multi-modal multi-task mixture of experts,” inThe Thirteenth International Conference on Learning Representations, 2025

    2025

  24. [24]

    Dynamical multimodal fusion with mixture-of-experts for localizations,

    B. Wang, Z. Shuai, F. Zhu, C. Huang, Y . Shen, Z. Zhang, Q. Yang, S. Muhaidat, and M. Debbah, “Dynamical multimodal fusion with mixture-of-experts for localizations,”arXiv preprint arXiv:2507.01337, 2025

    Pith/arXiv arXiv 2025

  25. [25]

    Graph multi-convolution and attention pooling for graph classification,

    Y . Xu, J. Wang, M. Guang, and C. Jiang, “Graph multi-convolution and attention pooling for graph classification,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    2024

  26. [26]

    A hierarchical mixture-of-experts frame- work for few labeled node classification,

    Y . Wang, Z. Yang, and X. Che, “A hierarchical mixture-of-experts frame- work for few labeled node classification,”Neural Networks, vol. 188, p. 107285, 2025

    2025

  27. [27]

    Mixture of experts for node classification,

    Y . Shi, Y . Wang, W. Liang, J. Zhang, P. Dong, and A. Li, “Mixture of experts for node classification,” inProceedings of the 2025 International Conference on Multimedia Retrieval, pp. 1154–1162, 2025

    2025

  28. [29]

    A multi-view mixture-of-experts based on language and graphs for molecular properties prediction,

    V . Y . Shirasuna, E. Soares, E. V . Brazil, K. F. A. Gutierrez, R. Cerqueira, S. Takeda, and A. Kishimoto, “A multi-view mixture-of-experts based on language and graphs for molecular properties prediction,” inICML 2024 AI for Science Workshop, 2024

    2024

  29. [30]

    Graph mixture of experts and memory-augmented routers for multivariate time series anomaly detec- tion,

    X. Huang, W. Chen, B. Hu, and Z. Mao, “Graph mixture of experts and memory-augmented routers for multivariate time series anomaly detec- tion,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 39, pp. 17476–17484, 2025

    2025

  30. [31]

    The more, the better? evaluating the role of eeg preprocessing for deep learning applications.,

    F. Del Pup, A. Zanola, L. F. Tshimanga, A. Bertoldo, and M. Atzori, “The more, the better? evaluating the role of eeg preprocessing for deep learning applications.,”IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2025

    2025

  31. [32]

    Future of alzheimer’s detection: Advancing diagnostic accuracy through the integration of qeeg and artificial intelligence,

    S. Rezaei, F. Asadirad, A. Motamedi, M. Kamran, F. Parsa, H. Samimi, P. Ghannadikhosh, M. Zahmatyar, S. A. Hosseinzadeh, and H. Arabi, “Future of alzheimer’s detection: Advancing diagnostic accuracy through the integration of qeeg and artificial intelligence,”NeuroImage, p. 121373, 2025

    2025

  32. [33]

    Using cnn saliency maps and eeg modulation spectra for improved and more interpretable machine learning-based alzheimer’s disease diagnosis,

    M. Lopes, R. Cassani, and T. H. Falk, “Using cnn saliency maps and eeg modulation spectra for improved and more interpretable machine learning-based alzheimer’s disease diagnosis,”Computational Intelli- gence and Neuroscience, vol. 2023, no. 1, p. 3198066, 2023

    2023

  33. [34]

    A novel method for diagnosing alzheimer’s disease using deep pyramid cnn based on eeg signals,

    W. Xia, R. Zhang, X. Zhang, and M. Usman, “A novel method for diagnosing alzheimer’s disease using deep pyramid cnn based on eeg signals,”Heliyon, vol. 9, no. 4, 2023

    2023

  34. [35]

    Deep learning into the future: Hybrid cnn-rnn for early detection of alzheimer’s disease,

    A. Kowshiga, T. Pavithra, V . Priyanka,et al., “Deep learning into the future: Hybrid cnn-rnn for early detection of alzheimer’s disease,” in2024 5th International Conference on Electronics and Sustainable Communication Systems (ICESC), pp. 940–946, IEEE, 2024

    2024

  35. [36]

    Fuzzy rnn model-based classification of alzheimer’s disease and dementia using brain eeg signals,

    J. Sreedhar, U. Pattapu, M. B. Reddy, S. Dara, K. K. Agrawal, P. Ku- mar, J. Ali, and A. Alkhayyat, “Fuzzy rnn model-based classification of alzheimer’s disease and dementia using brain eeg signals,”IEEE Transactions on Consumer Electronics, 2025

    2025

  36. [37]

    Adformer: A multi-granularity transformer for eeg-based alzheimer’s disease assessment,

    Y . Wang, N. Mammone, D. Petrovsky, A. T. Tzallas, F. C. Morabito, and X. Zhang, “Adformer: A multi-granularity transformer for eeg-based alzheimer’s disease assessment,”arXiv preprint arXiv:2409.00032, 2024

    Pith/arXiv arXiv 2024

  37. [38]

    Multi-frequency eeg and multi- functional connectivity graph convolutional network based detection method of patients with alzheimer’s disease,

    Y . Liu, L. An, H. Yang, and S. S. Ge, “Multi-frequency eeg and multi- functional connectivity graph convolutional network based detection method of patients with alzheimer’s disease,”Complex & Intelligent Systems, vol. 11, no. 8, pp. 1–17, 2025

    2025

  38. [39]

    Cognitive and neuropsychiatric correlates of eeg dynamic complexity in patients with alzheimer’s disease,

    A. C. Yang, S.-J. Wang, K.-L. Lai, C.-F. Tsai, C.-H. Yang, J.-P. Hwang, M.-T. Lo, N. E. Huang, C.-K. Peng, and J.-L. Fuh, “Cognitive and neuropsychiatric correlates of eeg dynamic complexity in patients with alzheimer’s disease,”Progress in Neuro-Psychopharmacology and Biological Psychiatry, vol. 47, pp. 52–61, 2013

    2013

  39. [40]

    Eeg2vec: Learning affective eeg rep- resentations via variational autoencoders,

    D. Bethge, P. Hallgarten, T. Grosse-Puppendahl, M. Kari, L. L. Chuang, O. Özdenizci, and A. Schmidt, “Eeg2vec: Learning affective eeg rep- resentations via variational autoencoders,” in2022 IEEE international conference on systems, man, and cybernetics (SMC), pp. 3150–3157, IEEE, 2022

    2022

  40. [41]

    Variational pathway reasoning for eeg emotion recognition,

    T. Zhang, Z. Cui, C. Xu, W. Zheng, and J. Yang, “Variational pathway reasoning for eeg emotion recognition,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 34, pp. 2709–2716, 2020

    2020

  41. [42]

    Variational instance-adaptive graph for eeg emotion recognition,

    T. Song, S. Liu, W. Zheng, Y . Zong, Z. Cui, Y . Li, and X. Zhou, “Variational instance-adaptive graph for eeg emotion recognition,”IEEE Transactions on Affective Computing, vol. 14, no. 1, pp. 343–356, 2021

    2021

  42. [43]

    Balancing active inference and active learning with deep variational predictive coding for eeg,

    A. Ofner and S. Stober, “Balancing active inference and active learning with deep variational predictive coding for eeg,” in2020 IEEE Interna- tional Conference on Systems, Man, and Cybernetics (SMC), pp. 3839– 3844, IEEE, 2020

    2020

  43. [44]

    Vsgt: variational spatial and gaussian temporal graph models for eeg-based emotion recognition,

    C. Liu, X. Zhou, J. Xiao, Z. Zhu, L. Zhai, Z. Jia, and Y . Liu, “Vsgt: variational spatial and gaussian temporal graph models for eeg-based emotion recognition,” inProceedings of the Thirty-Third International Joint Conference on Artificial Intelligence, pp. 3078–3086, 2024

    2024

  44. [45]

    An alternative model for mixtures of experts,

    L. Xu, M. Jordan, and G. E. Hinton, “An alternative model for mixtures of experts,”Advances in neural information processing systems, vol. 7, 1994

    1994

  45. [46]

    Empt: a sparsity transformer for eeg- based motor imagery recognition,

    M. Liu, Y . Liu, W. Shi, Y . Lou, Y . Sun, Q. Meng, D. Wang, F. Xu, Y . Zhang, L. Zhang,et al., “Empt: a sparsity transformer for eeg- based motor imagery recognition,”Frontiers in Neuroscience, vol. 18, p. 1366294, 2024. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 14

    2024

  46. [47]

    Eegmamba: Bidirectional state space model with mixture of experts for eeg multi-task classifica- tion,

    Y . Gui, M. Chen, Y . Su, G. Luo, and Y . Yang, “Eegmamba: Bidirectional state space model with mixture of experts for eeg multi-task classifica- tion,”arXiv preprint arXiv:2407.20254, 2024

    Pith/arXiv arXiv 2024

  47. [48]

    Mixture of experts for eeg-based seizure subtype classification,

    Z. Du, R. Peng, W. Liu, W. Li, and D. Wu, “Mixture of experts for eeg-based seizure subtype classification,”IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 31, pp. 4781–4789, 2023

    2023

  48. [49]

    Wavelet/mixture of experts network structure for eeg signals classification,

    E. D. Übeyli, “Wavelet/mixture of experts network structure for eeg signals classification,”Expert systems with applications, vol. 34, no. 3, pp. 1954–1962, 2008

    1954

  49. [50]

    Decoding the moving mind: Multi-subject fmri-to-video retrieval with mllm semantic grounding,

    X. Zhu, Y . Chai, R. Li, M. Lan, and L. Gao, “Decoding the moving mind: Multi-subject fmri-to-video retrieval with mllm semantic grounding,” bioRxiv, pp. 2025–04, 2025

    2025

  50. [51]

    Neuro-mobre: Exploring multi-subject multi-task intracranial decoding via explicit heterogeneity resolving,

    D. Wu, Y . Jia, S. Li, S. Zhao, J. Yang, and M. Sawan, “Neuro-mobre: Exploring multi-subject multi-task intracranial decoding via explicit heterogeneity resolving,”arXiv preprint arXiv:2508.04128, 2025

    Pith/arXiv arXiv 2025

  51. [52]

    Eeg emotion recognition via identity based multi-gate mixture-of-experts network,

    L. Yang, D. Liu, Q. Zhang, S. Chao, P. Ni, Q. Wang, and H. Sun, “Eeg emotion recognition via identity based multi-gate mixture-of-experts network,” in2022 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pp. 2498–2505, IEEE, 2022

    2022

  52. [53]

    Cognitmoe: A cognition-aware collaborative multi-expert network for bipolar disorder diagnosis,

    X. Zhu, Y . Wang, Y . Ma, Z. Zhang, and Y . Gao, “Cognitmoe: A cognition-aware collaborative multi-expert network for bipolar disorder diagnosis,”Neural Networks, p. 107854, 2025

    2025

  53. [54]

    Brainnet-moe: Brain-inspired mixture-of- experts learning for neurological disease identification,

    J. Zhang, X. Yu, T. Chen, C. Cao, M. Chen, Y . Zhuang, Y . Lyu, L. Zhang, L. Su, T. Liu,et al., “Brainnet-moe: Brain-inspired mixture-of- experts learning for neurological disease identification,”arXiv preprint arXiv:2503.07640, 2025

    Pith/arXiv arXiv 2025

  54. [55]

    Evaluating eeg com- plexity and spectral signatures in alzheimer’s disease and frontotemporal dementia: evidence for rostrocaudal asymmetry,

    K. Ghassemkhani, K. S. Saroka, and B. T. Dotta, “Evaluating eeg com- plexity and spectral signatures in alzheimer’s disease and frontotemporal dementia: evidence for rostrocaudal asymmetry,”npj Aging, vol. 11, no. 1, p. 50, 2025

    2025

  55. [56]

    Mgformer: A lightweight multi-granular transformer for subject-independent alzheimer’s classifi- cation,

    R. Rahman, A. K. al Azad, and S. Momen, “Mgformer: A lightweight multi-granular transformer for subject-independent alzheimer’s classifi- cation,”Biomedical Signal Processing and Control, vol. 110, p. 108230, 2025

    2025

  56. [57]

    Resting state eeg biomarkers of cognitive decline associated with alzheimer’s disease and mild cognitive impairment,

    A. H. Meghdadi, M. Stevanovi ´c Kari ´c, M. McConnell, G. Rupp, C. Richard, J. Hamilton, D. Salat, and C. Berka, “Resting state eeg biomarkers of cognitive decline associated with alzheimer’s disease and mild cognitive impairment,”PloS one, vol. 16, no. 2, p. e0244180, 2021

    2021

  57. [58]

    Auto-encoding variational bayes,

    D. P. Kingma and M. Welling, “Auto-encoding variational bayes,” in Proceedings of the 2nd International Conference on Learning Repre- sentations (ICLR), 2014. arXiv:1312.6114

    Pith/arXiv arXiv 2014

  58. [59]

    Adaptive mixtures of local experts,

    R. A. Jacobs, M. I. Jordan, S. J. Nowlan, and G. E. Hinton, “Adaptive mixtures of local experts,”Neural computation, vol. 3, no. 1, pp. 79–87, 1991

    1991

  59. [60]

    Graph classification by mixture of diverse experts,

    F. Hu, L. Wang, S. Wu, L. Wang, and T. Tan, “Graph classification by mixture of diverse experts,”arXiv preprint arXiv:2103.15622, 2021

    Pith/arXiv arXiv 2021

  60. [61]

    A dataset of scalp eeg recordings of alzheimer’s disease, frontotemporal dementia and healthy subjects from routine eeg,

    A. Miltiadous, K. D. Tzimourta, T. Afrantou, P. Ioannidis, N. Grigo- riadis, D. G. Tsalikakis, P. Angelidis, M. G. Tsipouras, E. Glavas, N. Giannakeas,et al., “A dataset of scalp eeg recordings of alzheimer’s disease, frontotemporal dementia and healthy subjects from routine eeg,” Data, vol. 8, no. 6, p. 95, 2023

    2023

  61. [62]

    Eegnet: a compact convolutional neural network for eeg-based brain–computer interfaces,

    V . J. Lawhern, A. J. Solon, N. R. Waytowich, S. M. Gordon, C. P. Hung, and B. J. Lance, “Eegnet: a compact convolutional neural network for eeg-based brain–computer interfaces,”Journal of neural engineering, vol. 15, no. 5, p. 056013, 2018

    2018

  62. [63]

    Vit2eeg: leveraging hybrid pretrained vision transformers for eeg data,

    R. Yang and E. Modesitt, “Vit2eeg: leveraging hybrid pretrained vision transformers for eeg data,”arXiv preprint arXiv:2308.00454, 2023

    Pith/arXiv arXiv 2023

  63. [64]

    Dual- transformer cross-attention framework for alzheimer’s disease detection via dpte-guided eeg channel selection and multi-modal integration,

    S. Y . Dharia, Q. Liu, S. D. Smith, and C. E. Valderrama, “Dual- transformer cross-attention framework for alzheimer’s disease detection via dpte-guided eeg channel selection and multi-modal integration,” Biomedical Signal Processing and Control, vol. 112, p. 108390, 2026

    2026

  64. [65]

    Deciphering bladder cancer-related circrna biomarkers: An ensemble model integrating deep learning and statistics for circrna analysis,

    Y . Ding, Y . Pan, R. C. Geyer, F. J. Vizeacoumar, F. S. Vizeacoumar, and F.-X. Wu, “Deciphering bladder cancer-related circrna biomarkers: An ensemble model integrating deep learning and statistics for circrna analysis,” inProceedings of the 15th ACM International Conference on Bioinformatics, Computational Biology and Health Informatics, pp. 1– 10, 2024

    2024

  65. [66]

    Graphmore: Mitigating topological heterogeneity via mixture of rie- mannian experts,

    Z. Guo, Q. Sun, H. Yuan, X. Fu, M. Zhou, Y . Gao, and J. Li, “Graphmore: Mitigating topological heterogeneity via mixture of rie- mannian experts,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 39, pp. 11754–11762, 2025

    2025

  66. [67]

    Moge: Mixture of graph experts for cross-subject emotion recognition via decomposing eeg,

    X.-H. Liu, W.-B. Jiang, W.-L. Zheng, and B.-L. Lu, “Moge: Mixture of graph experts for cross-subject emotion recognition via decomposing eeg,” in2024 ieee international conference on bioinformatics and biomedicine (bibm), pp. 3515–3520, IEEE, 2024

    2024

  67. [68]

    Mixture of weak and strong experts on graphs,

    H. Zeng, H. Lyu, D. Hu, Y . Xia, and J. Luo, “Mixture of weak and strong experts on graphs,” inThe Twelfth International Conference on Learning Representations, 2024

    2024

  68. [69]

    A novel approach to identify the brain regions that best classify adhd by means of eeg and deep learning,

    J. Sanchis, S. García-Ponsoda, M. A. Teruel, J. Trujillo, and I.-Y . Song, “A novel approach to identify the brain regions that best classify adhd by means of eeg and deep learning,”Heliyon, vol. 10, no. 4, 2024

    2024

  69. [70]

    American clinical neurophysiology society guideline 2: guidelines for standard electrode position nomenclature,

    J. N. Acharya, A. J. Hani, J. Cheek, P. Thirumala, and T. N. Tsuchida, “American clinical neurophysiology society guideline 2: guidelines for standard electrode position nomenclature,”The Neurodiagnostic Journal, vol. 56, no. 4, pp. 245–252, 2016

    2016

  70. [71]

    Resting-state eeg signatures of alzheimer’s disease are driven by periodic but not aperiodic changes,

    M. Kop ˇcanová, L. Tait, T. Donoghue, G. Stothart, L. Smith, A. A. Flores- Sandoval, P. Davila-Perez, S. Buss, M. M. Shafi, A. Pascual-Leone, et al., “Resting-state eeg signatures of alzheimer’s disease are driven by periodic but not aperiodic changes,”Neurobiology of Disease, vol. 190, p. 106380, 2024

    2024

  71. [72]

    Slowing of eeg background activity in parkinson’s and alzheimer’s disease with early cognitive dysfunction,

    N. Benz, F. Hatz, H. Bousleiman, M. M. Ehrensperger, U. Gschwandt- ner, M. Hardmeier, S. Ruegg, C. Schindler, R. Zimmermann, A. U. Monsch,et al., “Slowing of eeg background activity in parkinson’s and alzheimer’s disease with early cognitive dysfunction,”Frontiers in aging neuroscience, vol. 6, p. 314, 2014

    2014

  72. [73]

    Frontal white matter volume and delta eeg sources negatively correlate in awake subjects with mild cognitive impairment and alzheimer’s disease,

    C. Babiloni, G. Frisoni, M. Steriade, L. Bresciani, G. Binetti, C. Del Per- cio, C. Geroldi, C. Miniussi, F. Nobili, G. Rodriguez,et al., “Frontal white matter volume and delta eeg sources negatively correlate in awake subjects with mild cognitive impairment and alzheimer’s disease,” Clinical Neurophysiology, vol. 117, no. 5, pp. 1113–1129, 2006

    2006

  73. [74]

    Electroencephalographic rhythms in Alzheimer’s disease,

    R. Lizio, F. Vecchio, G. B. Frisoni, R. Ferri, G. Rodriguez, and C. Ba- biloni, “Electroencephalographic rhythms in Alzheimer’s disease,”In- ternational Journal of Alzheimer’s Disease, vol. 2011, no. 1, p. 927573, 2011

    2011

  74. [75]

    Quanti- tative eeg in the differential diagnosis of dementia subtypes,

    Y . Ol ˘gun, C. Aksoy Poyraz, M. Bozluolçay, and B. Ç. Poyraz, “Quanti- tative eeg in the differential diagnosis of dementia subtypes,”Journal of geriatric psychiatry and neurology, vol. 37, no. 5, pp. 368–378, 2024

    2024

  75. [76]

    Eeg-based minimum spanning tree analysis reveals network disruptions in alzheimer’s disease spectrum: An observational study,

    X. Ye, Y . Yan, Y . Wang, and J. Shi, “Eeg-based minimum spanning tree analysis reveals network disruptions in alzheimer’s disease spectrum: An observational study,”Frontiers in Aging Neuroscience, vol. 17, p. 1604345

  76. [77]

    Different oscillatory mechanisms of dementia-related diseases with cognitive impairment in closed-eye state,

    T. Zikereya, Y . Lin, Z. Zhang, I. Taguas, K. Shi, and C. Han, “Different oscillatory mechanisms of dementia-related diseases with cognitive impairment in closed-eye state,”Neuroimage, vol. 304, p. 120945, 2024

    2024

  77. [78]

    Evaluating brain electroencephalo- gram signal dynamics across cognitive disorders using information geometry,

    H. J. Choong, E.-j. Kim, and F. He, “Evaluating brain electroencephalo- gram signal dynamics across cognitive disorders using information geometry,”PLOS Complex Systems, vol. 2, no. 7, p. e0000059, 2025

    2025

  78. [79]

    Implication of eeg theta/alpha and theta/beta ratio in alzheimer’s and lewy body disease,

    K. Baik, J. H. Jung, S. H. Jeong, S. J. Chung, H. S. Yoo, P. H. Lee, Y . H. Sohn, S. W. Kang, and B. S. Ye, “Implication of eeg theta/alpha and theta/beta ratio in alzheimer’s and lewy body disease,”Scientific Reports, vol. 12, no. 1, p. 18706, 2022

    2022

  79. [80]

    Time- frequency functional connectivity alterations in alzheimer’s disease and frontotemporal dementia: An eeg analysis using machine learning,

    H. Zheng, H. Xiao, Y . Zhang, H. Jia, X. Ma, and Y . Gan, “Time- frequency functional connectivity alterations in alzheimer’s disease and frontotemporal dementia: An eeg analysis using machine learning,” Clinical Neurophysiology, vol. 170, pp. 110–119, 2025

    2025

  80. [81]

    Differences in quanti- tative eeg between frontotemporal dementia and alzheimer’s disease as revealed by loreta,

    K. Nishida, M. Yoshimura, T. Isotani, T. Yoshida, Y . Kitaura, A. Saito, H. Mii, M. Kato, Y . Takekita, A. Suwa,et al., “Differences in quanti- tative eeg between frontotemporal dementia and alzheimer’s disease as revealed by loreta,”Clinical Neurophysiology, vol. 122, no. 9, pp. 1718– 1725, 2011

    2011

This paper was first reviewed by deepseek-v4-flash on August 4, 2026.